Nuclear Physics B (Proc. Suppl.) 172 (2007) 327330 www.elsevierphysics.

com

Radiation hardness of CCD vertex detectors for the ILC

A. Sopczaka , K. Bekhoucheb , C. Bowderya , C. Damerellc , G. Daviesa , L. Dehimib , T. Greenshawd, M. Koziela , K. Stefanovc , T. Woolliscroftd , S. Wormc .
a b c

Lancaster University, UK LMSM Laboratory Biskra University, Algeria

CCLRC Rutherford Appleton Laboratory (RAL), UK Liverpool University, UK

Results of detailed simulations of the charge transfer ineciency of a prototype CCD chip are reported. The eect of radiation damage in a particle detector operating at the ILC is studied for two electron trap levels, 0.17 eV and 0.44 eV below the conduction band. Good agreement is found between simulations and an analytical model for the 0.17 eV level. Optimum operation is at about 250 K, approximately independent of readout frequency.

1. INTRODUCTION Any experiment exploiting the International Linear Collider (ILC) will require a detector to detect and measure short-lived particles but remain tolerant to radiation damage for its anticipated lifetime. This might contain chargecoupled devices (CCDs) which suer from both surface and bulk radiation damage. However, when considering charge transfer losses in buried channel devices only bulk traps are important. These defects create energy levels, hence electrons may be captured and later emitted back to the conduction band after a certain time. For a signal packet this may lead to a decrease in charge as it is transferred to the output and is quantied by its Charge Transfer Ineciency (CTI), where a charge of amplitude Q0 transported across m pixels will have a reduced charge given by Qm = Q0 (1 CTI)m . Previous studies of radiation-induced CTI have been reported [1]. The UK LCFI collaboration1 has been studying a device produced by e2V Technologies, with a manufacturers designation CCD58. It is a 2.1 Mpixel, three-phase buried-channel CCD with 12 m square pixels. Simulations of a simplied
1 LCFI

model of this have been performed with the ISETCAD package (v7.5), particularly the DESSIS program. It contains nine gates (numbered 1 to 9) which form the pixels. Each pixel consists of 3 gates but only one pixel is important for this studygates 5, 6 and 7. The overall length and depth of the simulated device are 44 m and 20 m respectively. The signal charge used in the simulation is chosen to be similar to the charge generated by a minimum ionising particle (MIP), corresponding to about 1620 electrons. This CTI study, at nominal clock voltage, focuses only on the bulk traps with energies 0.17 eV and 0.44 eV below the bottom of the conduction band. These will be referred to simply as the 0.17 eV and 0.44 eV traps. An incident particle with sucient energy is able to displace an atom from its lattice point leading eventually to a stable defect. These defects manifest themselves as energy levels. The trap concentrations and electron capture cross-sections used in the simulations are shown in table below. Et -Ec (eV) Type C (cm3 ) 0.17 Acceptor 1 1011 0.44 Acceptor 1 1011 (cm2 ) 1 1014 3 1015

stands for Linear Collider Flavour Identication

Each electron trap in the semiconductor material can either be empty (holding no electron) or full (holding one electron). In order to simulate

0920-5632/$ see front matter 2007 Elsevier B.V. All rights reserved. doi:10.1016/j.nuclphysbps.2007.08.064

328

A. Sopczak et al. / Nuclear Physics B (Proc. Suppl.) 172 (2007) 327330

the normal operating conditions of CCD58, partial trap lling was employed in the simulation (which means that some traps are full and some are empty) because the device will transfer many charge packets during continuous operation. 2. ANALYTICAL MODELS The following two simple analytical models are introduced to understand the underlying eects and to make comparisons with the DESSIS simulations (referred to as the full simulations). 2.1. Simple CTI Model Firstly, a simple analytical model is considered, based upon a single trapping levela so-called Simple CTI model. This is signicantly faster than a full simulation. It also provides a simple method to see the eect of changing parameters and demonstrates physics understanding. The charge transfer process is modelled by a dierential equation in terms of the dierent time constants and temperature dependence of the electron capture and emission processes. In the electron capture process, electrons are captured from the signal packet and each captured electron lls a trap. This occurs at the capture rate c . The electron emission process is described by the emission of captured electrons from lled traps back to the conduction band, and into a second signal packet at the emission rate e . The Shockley-Read-Hall theory [2] considers a defect at an energy Et below the bottom of the conduction band, Ec , and gives expressions for e and c . At low temperatures, the emission time constant e can be very large and of the order of seconds. The charge shift time is of the order of nanoseconds. A larger e means that a trap remains lled for much longer than the charge shift time. Further trapping of signal electrons is not possible and, consequently, the CTI is small at low temperatures. A peak occurs between low and high temperatures because the CTI is also small at high temperatures. This manifests itself because, at high temperatures, the emission time constant decreases to become comparable to the charge shift time. Now, trapped electrons rejoin their signal packet.

From the fraction of lled traps, a dierential equation can be derived where rf (t) = nt (t)/Nt is the time-dependent fraction of lled traps, nt (t) is the density of traps lled by electrons, and Nt is the density of traps. Considering that the traps are partially lled and using the initial condition: rf (0) = rf (tsh )etw /e where rf (0) is the fraction of lled traps after a mean waiting time, tw , the dierential equation can be solved to provide an expression for the CTI: 3Nt s rf (0) 1 etsh /s CT I = ns c
1 1 1 where s = c + e , and tsh is the shift-time. For one gate, tsh = 1/(3f), where f is the readout frequency. This denition is for CTI for a single trap level. The factor of three appears since there is a sum over the three gates that make up a pixel. The Simple Model has been adapted by including initially lled traps and by the incorporation of a so-called P factor to CTI:

P = etsh /e + e2tsh /e + e3tsh /e This models the situation where the trapped charge under gate 5 started to empty at time t minus three shift-times, that under gate 6 at t minus two shift-times and that under gate 7 at t minus one shift-time. An alternative factor, called P , has also been used to compare with simulated data. This is dened as: P = 1 + etsh /e + e2tsh /e and models the situation one shift-time earlier than for P . 2.2. Improved Model The second analytical model is the Improved Model (IM), based on the work of T. Hardy et al. [3]. It is improved by adjusting initial assumptions to t the study of CCD58 and also considers the eect of a single trapping level, but only includes the emission time in its dierential equation where nt is the density of lled traps. The traps are initially lled for this model and tsh . Nevertheless, to be consistent with c the full DESSIS simulations (that use partially lled traps) the Improved Model uses a time constant between the lling of the traps such that the

A. Sopczak et al. / Nuclear Physics B (Proc. Suppl.) 172 (2007) 327330

329

traps remain partially lled when the new electron packet passes through the CCD. The solution of this dierential equation leads to another estimator of the CTI: 3Nt tjoin /e e CT I I = 1 etsh /c etemit /e ns where temit = tw is the total emission time from the previous packet, and tjoin is the time period during which the charges can join the parent charge packet. 3. SIMULATION RESULTS The CTI dependence on temperature and readout frequency was explored using DESSIS simulations. Figure 1 shows the CTI for simulations with partially lled 0.17 eV traps at dierent frequencies for temperatures between 123 K and 260 K, with a nominal clock voltage of 7 V. Figure 2. CTI for simulations with 0.44 eV partially lled traps clocked at 7, 25 and 50 MHz.

cause only a negligible CTI at temperatures lower than 250 K due to the long emission time and thus traps remain fully lled at lower temperatures (Fig. 2).

x 10

Simple model PFactor model P Factor model Improved Model Full simulation 2

CTI 1

Figure 1. CTI for simulations with 0.17 eV partially lled traps clocked at 7, 25 and 50 MHz.

0 120

140

160

180 200 Temperature (K)

220

240

260

The CTI increases as frequency decreases. For higher readout frequency there is less time to trap the charge, thus the CTI is reduced. At high temperatures the emission time is so short that trapped charges rejoin the passing signal. Simulations were also carried out with partially lled 0.44 eV traps at temperatures ranging from 250 K to 500 K. This is because previous studies on 0.44 eV traps have shown that these traps

Figure 3. CTI values against temperature for different models with 50 MHz readout. See text. The peak CTI is higher for lower frequencies with little temperature dependence of the peak position. The optimal operating temperature is about 250 K. Figure 3 shows that the basic Simple Model does not agree well with the full simulation. Applying the P factor appears to over-

330

A. Sopczak et al. / Nuclear Physics B (Proc. Suppl.) 172 (2007) 327330

compensate for the deciencies and the P factor gives a reasonable but not perfect agreement. Figure 4 compares the full DESSIS simulation for 0.17 eV and 0.44 eV traps and clocking frequency of 50 MHz to the Improved Model. It emphasises the good agreement between the model and full simulations at temperatures lower than 250 K with 0.17 eV traps, but shows a disagreement at higher temperatures for the 0.44 eV traps.

analysed. The results conrm the dependence of CTI with the readout frequency. At low temperatures (< 250 K) the 0.17 eV traps dominate the CTI, whereas the 0.44 eV traps dominate at higher temperatures. Good agreement between simulations and a socalled Improved Model has been found for 0.17 eV traps but not for 0.44 eV traps. This shows the limitations of the Improved Model with respect to the full simulation. The optimum operating temperature for CCD58 in a high radiation environment is found to be about 250 K. ACKNOWLEDGMENTS This work is supported by the Particle Physics and Astronomy Research Council (PPARC) and Lancaster University. The Lancaster authors wish to thank Alex Chilingarov, for helpful discussions, and the particle physics group at Liverpool University, for the use of its computers. AS would like to thank the organizers of the conference for their hospitality.

Figure 4. CTI values for simulations for 0.17 eV versus 0.44 eV partially lled traps at clocking frequency 50 MHz. Comparison of Improved Model (IM) with full DESSIS simulation. If the 0.44 eV trap electron capture crosssection in the Improved Model is increased to 1014 cm2 , a somewhat better agreement is found. However it is clear that there are limitations with the Improved Model. They could relate to a breakdown of the assumptions at high temperatures, to ignoring the precise form of the clock voltage waveform, or to ignoring the pixel edge eects. Further studies are required. 4. CONCLUSIONS The Charge Transfer Ineciency (CTI) of a CCD device has been studied with a full simulation and compared with analytical models. Partially lled traps from the 0.17 eV and 0.44 eV trap levels have been implemented in the full simulation and variations of the CTI with respect to temperature and frequency have been

REFERENCES 1. J.R. Janesick, Scientic Charge-Coupled Devices, SPIE Press, ISBN 0819436984 (2001); K. Stefanov et al., IEEE Trans. Nucl. Sci., 47 (2000) 1280; O. Ursache, Diploma thesis, University of Siegen (Germany), 2003; J.E. Brau, O. Igonkina, C.T. Potter and N.B. Sinev, Nucl. Instr. and Meth. A549 (2005) 117; J.E. Brau and N.B. Sinev, IEEE Trans. Nucl. Sci. 47 (2000) 1898; A. Sopczak, IEEE 2005 Nuclear Science Symposium, San Juan, USA, Proc. IEEE Nuclear Science Symposium Conference Record, N377 (2005) 1494. 2. W. Shockley and W.T. Read, Phys. Rev. 87 (1952) 835; R.N. Hall, Phys. Rev., 87 (1952) 387. 3. T. Hardy, R. Murowinski and M.J. Deen, IEEE Trans. Nucl. Sci. 45 (1998) 154.